References

Recent atmospheric studies have evidenced the imprint of large N 2 O sources in tropical/subtropical lands. This source might be attributed to agricultural areas as well as to natural humid ecosystems. The uncertainty related to both sources is very high, due to the scarcity of data and low frequency of sampling in tropical studies, especially for the African continent. The principal objective of this work was to quantify the annual budget of N 2 O emissions in an African tropical rain forest. Soil N 2 O emissions were measured over 19 months in Ghana, National Park of Ankasa, in uphill and downhill areas, for a total of 119 days of observation. The calculated annual average emission was 2.33 ± 0.20 kg N-N 2 O ha −1 yr −1 , taking into account the proportion of uphill vs. downhill areas, the latter being characterized by lower N 2 O emissions. N 2 O fluxes peaked between June and August and were significantly correlated with soil respiration on a daily and monthly basis. No clear correlation was found in the uphill area between N 2 O fluxes and soil water content or rain, whereas in the downhill area soil water content concurred with soil respiration in determining N 2 O flux variability. The N 2 O source strength calculated in this study is very close to those reported for the other two available studies in African rain forests and to the estimated mean derived from worldwide studies in humid tropical forests (2.81 ± 2.02 kg N-N 2 O ha −1 yr −1 ).


Introduction
Tropical forests are a key ecosystem for terrestrial carbon cycling, being the most productive ecosystems on earth (Grace et al., 2001) and accounting for 59 % of the global carbon pool in forests (Dixon et al., 1994).Recent evidence indicates that these ecosystems might have a key role not only in the C cycle but also in the global atmospheric balance of the greenhouse gas nitrous oxide (N 2 O).Kort et al.'s (2011) measurements of vertical atmospheric profiles (from surface to 14 km altitude) between 67 • S to 85 • N latitude have shown significant sources of N 2 O in tropical areas with high temporal variability.The available data to support these atmospheric observations are still quite limited in terms of geographical coverage, length of sampling campaign and significance of spatial replication.Most of the available data come from Brasil (Keller et al., 1983, Keller et al., 1988;Verchot et al., 1999;Maddock et al., 2001;Melillo et al., 2001;Davidson et al., 2004;Keller et al., 2005;Neto et al. 2010) and Central America (Matson and Vitousek, 1987;Keller et al., 1993;Keller and Reiners, 1994;Weitz et al., 1998).Only very recently have some data from tropical ecosystems in Asia been published (Verchot et al., 2006;Werner et al., 2006;Yan et al., 2008).Few observations are available for Oceania (Breuer et al., 2000;Kiese and Butterbach-Bahl, 2002;Kiese et al., 2003) and just two studies are available for tropical rain forests in the African continent (Serca et al., 1994;Werner et al., 2007a).Moreover, most of the available studies in remote tropical areas are often based on short campaigns of a few days of measurements, which make the annual estimate highly uncertain.Given the wide surface covered by tropical forests, there is a clear need to have a sufficient number of studies to improve our estimates and to help to validate theoretical estimates and downscaling quantifications (Potter et al., 1996;Stehfest and Bouwman, 2006;Werner et al., 2007b).Africa is, in this respect, the least-investigated continent.
The principal objective of this work was to quantify the annual budget of N 2 O emissions in an African tropical rain forest.N 2 O fluxes were sampled monthly, with a frequency of 6 days per month, for 19 months.To better understand the dynamics of N 2 O production, soil CO 2 emissions, soil temperature and soil water content were also monitored and their variability related to N 2 O flux variability.

The study sites
The study site (05 • 16 11 N, 02 • 41 41 W), located in Ankasa Wildlife Protected Area, is an ancient rainforest with the highest biodiversity in Ghana.The forest has an area of about 500 km 2 and in 1976 became a wildlife protected area, to preserve the forests from cutting, slashing and burning, and conversion to agro-ecosystems.A core area of about 343 km 2 represents the only protected rain forest area, in an almost pristine state, in Ghana.The mean annual temperature is about 25 • C and the mean annual precipitation is between 1500-2000 mm, depending on the year, mainly concentrated from March to mid-July and from September to November, with a relative humidity ranging from 75 % to 90 %.A relatively dry period generally occurs between December and February (Hall and Swaine, 1981).The landscape is characterized by the presence of hills with an average elevation of 90 m a.s.l., alternating with steep valleys.Soils are deeply weathered and highly acidic and are classified as Oxisols (Ahn, 1961).Soil, developed on coherent biotite-rich granites of the Cape Coast complex, forms the Ankasa association, consisting of "Abenia" series, the most widespread series on top of the hills, which alternates to the "Ankasa" series, along the slopes.Alluvial soils occupy the bottom of the fluvial valleys.The main characteristics of the soil sampled in uphill and downhill study areas are presented in Table 1.Surface litter C in the hill top area is 15 ± 9 Mg C ha −1 , and a preliminary estimate of the aboveground biomass, including live and dead wood, is 138-170 Mg C ha −1 (Chiti et al., 2010).

Experimental set up
Soil N 2 O and CO 2 flux measurements were carried out during year 2009 and 2010.Sampling started in 2009 in an uphill area where a tower for eddy covariance measurements is located.The micrometeorological station was identified as the center of 4 quadrants (North-east, East-south, Southwest, West-north) and 2 chambers were positioned at a minimum distance of 50 m from the tower, in each quadrant (8 replicates in all).A second plot of about one hectare was set up in 2010 at the bottom of the valley, where soil conditions are expected to change significantly, in particular soil humidity.Also in this case 8 replicates were used.Chamber position was changed every day to improve the coverage of spatial variability, taking care to insert the chamber base in the ground about 3 h before starting the measurements so as to avoid false fluxes caused by soil pressure changes and to avoid effects of root mortality and decomposition which might occur in the first weeks following collar insertion into the soil (Keller et al. 2000).Replicates were placed at a minimum distance of 10 m apart.Within each plot, monthly campaigns of soil-atmosphere gas exchange were carried out, over six consecutive days.Sampling in the uphill area started in April 2009 while in the downhill area sampling started in May 2010, data are reported up to November 2010.The initial experimental set-up included soil temperature and water content measurements undertaken by local micrometeorological station.However, a failure of sensors due to lightning hampered the quality of data.Therefore, from May 2010, soil temperature (HI93510 thermometer, Hanna Instruments Canada Inc., Laval, Quebec) and volumetric soil water content (ThetaProbe ML2, Delta-T Device Ltd, Cambridge, UK) were hand-measured adjacent to each chamber (0-5 cm depth), 5 cm from the chamber edge, at each sampling date.Gas sampling was generally done between 10 and 12 a.m.Soil temperature has an annual average maximal daily variation of about 0.5 • C at 5 cm depth and 0.1 • C at 10 cm depth (meteorological station at site, 2008-2009 data).Soil water content was expressed as water-filled pore space % (WFPS) equal to 100×θ v )/ε, where θ v is the volumetric water content and ε is the total porosity (ε = 1 -bulk density/particle density).

Gas flux determination
Gas fluxes were measured using closed static chambers (Hutchinson and Mosier, 1981;Smith et al., 1995) made of PVC collars (7 cm high with a diameter of 15 cm), inserted in the soil to about 5 cm depth, and chamber lids (20 cm high with a diameter of 15 cm).To determine gas fluxes, gas (30 ml) was sampled, using gas-tight syringes, from the chamber headspace at 0, 30 and 60 min after closure; it was immediately stored in pre-evacuated gas-tight vials (20 ml), slightly overpressurized, which were sealed with thermal glue and shipped for gas chromatographic analysis (TRACE GC ULTRA, THERMO SCIENTIFIC).N 2 O flux rates were determined by linear regression of the three sampling points for each chamber and by applying a temperature and pressure correction.The analytical precision of the GC for standards at ambient concentration was approximately 3 %, using one standard deviation as a measure of a Determined using the pipette method after destruction of the organic cement using sodium hypochlorite at pH 9 (Mikutta et al., 2005).b Bulk density determined with the core method (Blake and Hartge 1986).c Measured in deionised water with a sure-flow electrode, using a ratio soil-solution of 1:2.5 (w/w).
mean error.The analytical set up (Loftfield et al., 1997) allowed both N 2 O and CO 2 fluxes to be measured from the same chamber.Gas was loaded on a 2 ml loop connected to a 10-port valve (Valco Europe, Switzerland).A pre-column of 1 m (O.D. 1/8 , 0.08 I.D.), filled with Porapak 80-100 Q and maintained at 60 • C, was connected to the 10-port valve in order to operate front-flush and back-flush.From the precolumn, the gas passed into the main column (T Porapak 80-100 Q, O.D. 1/8 , 0.08 I.D., 2 m length), also held at 60 • C and was then directed to an electron capture detector (ECD) held at 280 • C for the determination of CO 2 and N 2 O concentration.Pure nitrogen was used as carrier gas at a flow rate of 40 cm 3 min −1 .A test was made to evidence possible interference of CO 2 with N 2 O concentration determination (Wang et al., 2010), by using two N 2 O standards (320 ppb and 5000 ppb) which were mixed with CO 2 at different concentrations (range 1000-10000 ppm), no significant effect of CO 2 concentration on the peak area of N 2 O was found.Four concentrations of calibrated standards were used and were injected in duplicate every 20 samples to allow for monitoring instrument drift.The CO 2 fluxes determined using this approach can be used for comparative analysis between sites and to identify trends in soil respiration (Yan et al., 2009;Mapanda et al., 2010;Liu et al., 2011) but are not suitable for precisely quantifying soil CO 2 efflux and site budgets, due to the length of incubation time which might bring to flux underestimation.For this latter goal CO 2 efflux are usually determined by an infrared gas analyzer (IRGA) using dynamic closed or open chambers (Rayment and Jarvis, 1997;Heinemeyer and McNamara, 2010).For this reason no attempt was made in this study to quantify the annual budget of soil CO 2 emissions.

Statistics
A two-way analysis of variance was used to compare gas fluxes, soil temperature and water-filled pore space measured in the downhill and uphill plots.A normality test (Kolmogorov-Smirnov, with Lilliefors correction, Sigma Stat, Jandel Scientific) was performed before running parametric tests (Sigma Stat, Jandel Scientific).When the difference was significant (P <0.05) an "all pairwise" comparison was carried out using the "Student Newman-Keuls test".When normality tests failed, a Kruskal-Wallis ANOVA on ranks was performed.Simple linear regressions, multiple linear regression and non-linear regression analysis were performed to find the relationship between independent and dependent variables (Sigma Stat, Jandel Scientific).Significant differences were at the P <0.05 level.

Spatial and temporal variability of N 2 O fluxes
Single-chamber N 2 O fluxes varied between -0.15 and 29.13 mg N 2 O m −2 d −1 in the uphill area and -0.53 and 16.62 mg N 2 O m −2 d −1 in the downhill area.On 46 % of sampling days, the daily spatial variability (range CV % 33-288) was higher than the annual temporal flux variability calculated on daily averages (CV % 110).N 2 O flux showed a marked seasonal variability (Fig. 1a), with daily average fluxes ranging from 0.04 to 8.04 mg N 2 O m −2 d −1 in the uphill area and from 0.06 to 2.84 mg N 2 O m −2 d −1 in the downhill area (Fig. 1a).In both environments, the distribution of daily mean values was lognormal.
Referring to the period when both environments were monitored, the average daily N Comparing the downhill and uphill N 2 O cumulative fluxes over the same period of observation (May to November 2010), the total downhill N 2 O emission was only 44 % of the emission measured in the uphill area.The downhill surface was estimated to be about 20 % of the total surface area of the Ankasa National Park, derived by SRTM DEM at 90 m resolution (CGIAR -CSI free source).The calculated weighted annual emission per hectare, taking into account the uphill/downhill ratio, was 2.33 ± 0.20 N-N 2 O ha −1 yr −1 , a value which is within the error estimate of the calculated flux for uphill areas.

Relationship between environmental drivers and N 2 O fluxes
The potential drivers of N 2 O fluxes measured at site were soil temperature, soil water content, expressed as soil waterfilled pore space, soil CO 2 fluxes, monthly air temperature and monthly rainfall, calculated as the average of data from the two weather stations closest to the Ankasa site (TRMM database).The seasonal trend of N 2 O emissions was similar to that of soil respiration (Fig. 1a, b).On a daily basis, higher fluxes of N 2 O emissions corresponded to higher soil respiration rates (Fig. 2).On a monthly basis, the match between temporal trends of the two gases was particularly clear (Fig. 3a, b).The wider seasonal variations of N 2 O emissions, compared with soil respiration, most probably reflected the complex interplay of environmental factors which concurred in N 2 O production and emission.Mean monthly trend values of N 2 O and CO 2 emissions did not strictly correspond with trends of average air temperature and total rainfall (Fig. 3c).On a daily basis, variability of N 2 O and CO 2 fluxes were related to variations of soil water-filled pore space and soil temperature.For the uphill area, a lognormal relationship was found between both average daily N 2 O emissions and soil respiration and WFPS (Fig. 4a), with a gas flux peak between 30 % and 35 % WFPS, slightly shifted towards higher WFPS values for N 2 O compared to CO 2 fluxes (Fig. 4a,  b).In the downhill area no significant trend was found between N 2 O or CO 2 fluxes and WFPS, although both gases showed a tendency to increase for WFPS values above 50 %.However, a significant relationship was found between N 2 O and a linear combination of CO  daily N 2 O and CO 2 fluxes significantly increased in a range of soil temperatures comprised between 22.5 and 25 • C, the increase being less marked for CO 2 fluxes (Fig. 5).The narrow range of temperatures in which a sharp increase of N 2 O flux was observed corresponded to intermediate values of soil WFPS (Fig. 4a).Multilinear regression analysis did not evidence any other significant relationship between analyzed variables and N 2 O emissions.

Discussion
The annual budget calculated for the Ankasa rain forest (2.33 ± 0.20 kg N-N 2 O ha −1 yr −1 ) is very close to the annual N 2 O emission estimates reported by Serca et al. (1994) for a primary rain forest of the Mayombe region in the Congo (2.9 kg N-N 2 O ha −1 yr −1 ) and by Werner et al. (2007a) for a mountain rain forest in the Kakamega forest national park in Kenya (2.6 kg N-N 2 O ha −1 yr −1 ).The three African sites, including Ankasa, have similar rain regime and acidic soil; Ankasa is characterized by a higher percentage of sand in the soil texture, compared with the other two sites, the Kenya mountain forest has the highest content of soil C (about twice Ankasa values, in the top 5 centimeters), whereas Ghana and Congo soil C contents are comparable.Although the observed differences of soil C and texture might be expected to influence the N 2 O emission rates (Kiese et al., 2005;Werner et al., 2007b), the observed small difference of annual N 2 O emissions among sites indicate that the effect of such influence most probably was within the uncertainty associated to the annual N 2 O emission estimate.On the other  2007a) also pointed out that while short-term variations of N 2 O might be associated to significant changes in soil moisture, hence to isolated rain events, seasonal N 2 O emission patterns and the annual budget mainly depend on decomposition rates which control N availability.These observations are in accordance with our results, which indicated a very similar pattern of N 2 O emissions and soil respiration on a monthly average basis and a good correlation between the two variables at daily time step.In the well-drained uphill site, where WFPS never exceeded 50 %, soil respiration was the independent factor, the variability of which best predicted the variability of N 2 O emissions.Soil CO 2 emission, namely soil respiration, is the result of combined rates of autotrophic and heterotrophic respiration.The partitioning of these two components in the overall CO 2 soil efflux is variable depending on site characteristics, time of the year and of the day (Sotta et al., 2004).Both these components tend to increase with increasing soil temperature and moisture, although in tropical soil CO 2 efflux has been reported to be depressed by strong rain events due to both the limiting of gas diffusion and the decrease of soil temperature (Sotta et al., 2004).Although we could not distinguish between autotrophic and heterotropic sources of soil CO 2 emissions, we might expect that both processes would have influenced N 2 O production positively.High respiration rates, no matter which organism is involved, tend to decrease soil oxygen content and favor the development of anaerobic hotspots of N 2 O production within aerobic soils (Smith et al., 1990).Higher heterotrophic respiration rates can be expected to coincide with peaks of fresh organic matter input (litterfall), which also represent a source of N for microbes.In rain forests, peaks of litterfall have been reported to occur mostly at the onset of the rainy season (Muoghalu et al. 1993;De Moraes et al., 1999).Considering that air humidity inside the close and shaded canopy of a rain forest is always quite high, humidity and temperature conditions can generally be considered not limiting for decomposition in the humid tropics, compared with temperate and boreal forest ecosystems, so that a flow of N which supports nitrification and denitrification processes and N 2 O emissions (Firestone and Davidson, 1989) is assured during the whole year.An increase of decomposition rates and mineralization can be expected to coincide with peaks of litter fall and in the first weeks following the onset of heavy precipitation events if they are close to peaks of litter fall (Werner et al., 2007a).The lower fluxes measured in the downhill area, which coincided with much higher WFPS (37-78 %) compared with uphill, might in part be explained by lower soil respiration rates but also by a predominance of reducing conditions during part of the year, leading to further reduction of produced N 2 O to N 2 in soil hotspots where WFPS is close to saturation.The occurrence of such hotspots of anaerobic microbial activity is also confirmed by measured CH 4 fluxes which showed high emission rates only in the downhill area (Castaldi et al., 2013).The average annual N 2 O emission estimate derived from the three available studies in African rain forests (Ghana, Kenya, Congo), 2.6 ± 0.3 kg N-N 2 O ha −1 yr −1 , is quite close to the global average calculated for tropical humid forests from different continents, 2.81 ± 2.02 kg N-N 2 O ha −1 yr −1 (Table 2), excluding studies where estimates are based on sampling campaigns carried out only in one season (dry or wet).Of the studies reported in Table 2, which represent the most cited literature for humid tropical forests for which a tentative site assessment of N 2 O emission strength is made, about 50 % are characterized by medium to high uncertainty, in particular concerning the temporal frequency of sampling.Only nine of the 41 presented studies show an average sampling frequency higher than 4 samples per month.Within the group of sites characterized by medium/low uncertainty e spatial uncertainty classified as high (H) when no chambers <5, medium (M) when 5<no.ch <10, low (L) when no. ch >10; temporal uncertainty classified as high (H) when only one season is sampled or when over the year we have less than 1 measurement per month, medium (M) when sampling frequency is between 1 and 4 days per month, low (L) when sampling frequency is higher than 4 days per month.
the range of observed fluxes varies from 0.94 to 7.45 kg N-N 2 O ha −1 yr −1 .All the sites which show fluxes higher than 3 kg N-N 2 O ha −1 yr −1 are characterized by annual total rainfall higher than 2000 mm, with the sole exception of Kauri Creek site in Australia (Table 2), hence higher than rainfall at Ankasa site.Rainfall represents both a proximal and distal factor which exert a positive control (stimulation) on N 2 O emission, because it influences soil water-filled pore space, NPP, and decomposition rates.Moreover, at increasing rainfall rates there is a progressive reduction of the length of periods with clear dry conditions, which are less favorable for N 2 O production.
www.biogeosciences.net/10/4179/2013/Biogeosciences, 10, 4179-4187, 2013 Results showed that the rain forest in Ghana has annual emission rates very close to other rain forest sites of Central Africa characterized by similar rainfall conditions.Fluxes were characterized by high temporal variability, which supported the extreme importance of adequate sampling frequency in this natural environment.Moreover, the study also showed the importance of spatial coverage, with valley areas having statistically different fluxes from uphill areas.This is a common feature of many tropical landscapes which need to be taken into account when scaling up.

Fig. 1 .
Fig. 1.Daily mean values (± one st dev) of N 2 O fluxes (a) and soil respiration (b) measured in the uphill and downhill areas in 2009 and 2010.

Fig. 2 .
Fig. 2. N 2 O daily average fluxes plotted versus daily average CO 2 fluxes measured in the uphill and downhill sites.Plain line is a linear fit of data (R 2 = 0.57, P <0.0001, DF = 58).

Fig. 3 .
Fig. 3. Monthly mean N 2 O flux (bars) and soil CO 2 flux (black dots) calculated for the uphill (a) and downhill (b) areas; mean air temperature and total monthly rainfall (c).

Table 1 .
Main chemico-physical characteristics (± 1 st dev, n = 6) of the soil sampled in the study areas.

Table 2 .
Davidson et al 2004;ates (italic) of N 2 O emission (kg N ha −1 yr −1 ) as reported by authors or calculated (where indicated) from literature studies in humid tropical forests.Reported campaigns occurred only in dry (D) or wet (W) seasons; over a whole year (Y); during periods longer than one year (months are indicated).Arbitrary uncertainty levels are given (explained in Tablenotes) not as an absolute level of uncertainty but as a comparative indication among studies.samesiteofDavidson et al 2004;b calculated from annual average; c extrapolated from monthly measurements; d extrapolated from mean estimates for the sampling period; a